FDA premarket cybersecurity rules start with section 524B of the FD&C Act: if your medical device runs software and can connect to anything, it is probably a “cyber device”, and your submission must include three things: a plan to find and fix postmarket vulnerabilities, processes that give reasonable assurance the device is cybersecure (including postmarket updates and patches), and a software bill of materials. FDA’s February 2026 premarket cybersecurity guidance turns those three duties into documents: a threat model, a risk assessment, an SBOM with support status, security architecture views, test reports, labeling and a cybersecurity management plan. For the firmware team, that means a signed update path that can recover from failure, security controls you can point to in code, and test evidence that traces back to each one.
This article reads the statute and the current guidance and translates them into firmware work and records. It is not regulatory or legal advice. Your regulatory lead owns the submission strategy, and FDA’s own documents, linked throughout, are the authority.
Fumbar is a Denver firm that writes embedded firmware and over-the-air update systems for hardware startups. We are not a regulatory consultancy and we do not prepare submissions; we write firmware and the engineering records that go into them.
What section 524B requires, in the statute’s words
Section 524B has four requirements and one definition, and they are short enough to read in full at 21 U.S.C. 360n-2. It was added by the Food and Drug Omnibus Reform Act of 2022, enacted December 29, 2022, and took effect 90 days later, on March 29, 2023.
The sponsor of a submission for a cyber device shall:
- “submit to the Secretary a plan to monitor, identify, and address, as appropriate, in a reasonable time, postmarket cybersecurity vulnerabilities and exploits, including coordinated vulnerability disclosure and related procedures”;
- “design, develop, and maintain processes and procedures to provide a reasonable assurance that the device and related systems are cybersecure, and make available postmarket updates and patches to the device and related systems”, on “a reasonably justified regular cycle” for known unacceptable vulnerabilities and “as soon as possible out of cycle” for critical vulnerabilities that could cause uncontrolled risks;
- “provide to the Secretary a software bill of materials, including commercial, open-source, and off-the-shelf software components”; and
- comply with any further requirements FDA sets by regulation.
Here is what each clause asks of the people who write the firmware:
| Clause | Plain reading | Firmware work behind it |
|---|---|---|
| 524B(b)(1) plan | You will notice, assess and fix vulnerabilities after launch, and accept reports from outsiders | Version reporting from the field, crash and security event logs, a way to reproduce any shipped build |
| 524B(b)(2) processes and patches | The device is designed securely, and you can deliver fixes on a schedule and in an emergency | Secure boot, signed updates, anti-rollback, recovery from failed updates, a release process that can ship a security-only build |
| 524B(b)(3) SBOM | You know every software component in the device | A build that emits a machine-readable component list, including vendor SDKs and binary blobs |
The requirements apply to 510(k) submissions (including Special and Abbreviated), PMAs and PMA supplements, Product Development Protocols, De Novo requests and Humanitarian Device Exemptions, according to FDA’s FAQ. The same FAQ notes that FDA’s refuse-to-accept transition policy for cyber devices expired on October 1, 2023, and that a 510(k) sent through eSTAR is put on a technical screening hold if its cybersecurity section lacks accurate responses and attachments.
Is your device a cyber device?
A device is a cyber device if it meets all three tests in section 524B(c): it “includes software validated, installed, or authorized by the sponsor as a device or in a device”, it “has the ability to connect to the internet”, and it contains such characteristics “that could be vulnerable to cybersecurity threats.” FDA reads every part of that broadly, so most connected startup devices qualify.
On software, FDA’s February 2026 guidance says a cyber device includes devices that “are or contain software, including software that is firmware or programmable logic.” A microcontroller running your code meets the first test.
On connectivity, FDA counts devices that can reach the internet “whether intentionally or unintentionally, through any means.” Its illustrative list includes network, server or cloud connections; radio links such as Wi-Fi, cellular, Bluetooth and Bluetooth Low Energy; magnetic inductive communications; and hardware connectors such as USB, Ethernet and serial ports. A footnote makes the point sharply: a device serviced over USB, even briefly, “is therefore considered to have the ability to connect to the internet.”
If you are unsure, FDA’s FAQ says you may contact the agency. Two further points from the same FAQ matter even if you conclude you are not a cyber device:
- “The premarket guidance recommendations are applicable to all medical devices that have or are software.” Section 524B adds legal requirements on top; the guidance applies either way.
- “Cybersecurity protection by design is not optional.”
Example (fictional). Juniper Vitals, a two-founder Denver startup, is building the JV-2, a skin-temperature patch for patients recovering at home after surgery. The patch runs firmware on a Bluetooth LE microcontroller, streams readings to a phone app, and the app forwards them to a cloud dashboard that clinicians watch. The firmware is the founders’ own code on Zephyr with the MCUboot bootloader. The patch has no internet stack of its own, but it talks Bluetooth LE to a phone that does, and it has a serial service port used at the factory. Under FDA’s reading, both count, so the JV-2 is a cyber device and its 510(k) will need everything in section 524B(b).
Which guidance is current, and what changed in February 2026
The current document is “Cybersecurity in Medical Devices: Quality Management System Considerations and Content of Premarket Submissions”, final guidance dated February 2026. FDA’s guidance page says it “supersedes the final guidance … issued June 27, 2025.” Summaries written before February 2026 describe the June 2025 version, so check the date on anything you rely on.
The guidance’s own history table explains the change. The February 2026 revisions were issued under Level 2 procedures and align the text with the amendments to 21 CFR 820, the Quality Management System Regulation (QMSR), which incorporates ISO 13485 by reference. The current text cites ISO 13485 subclauses for design and development, risk management and corrective action. FDA’s cybersecurity page dates the issue February 2, 2026, and notes that the June 27, 2025 version added Section VII on cyber devices, which carries forward.
Two reading rules help. First, the guidance itself says FDA guidance documents “do not establish legally enforceable responsibilities” and that “should” means recommended, “unless specific regulatory or statutory requirements are cited.” Section 524B is such a requirement, so in Section VII the words “must” and “required” mean what they say. Second, FDA’s 2016 postmarket cybersecurity guidance still applies after launch, and the premarket guidance leans on its distinction between controlled and uncontrolled risk.
The documents FDA expects, and the firmware behind each
FDA recommends a security risk management report as the backbone of the submission, built from a threat model, a cybersecurity risk assessment, an SBOM with support information, a vulnerability assessment and an assessment of unresolved anomalies, all traceable to each other and to testing. Most of these documents describe firmware decisions, so the firmware team writes or feeds a large share of them.
| Document (guidance section) | What it must show | What the firmware team supplies |
|---|---|---|
| Threat model (V.A.1) | Risks and mitigations across the whole system, with stated assumptions | Interfaces, protocols, boot chain, debug access and update path as built |
| Cybersecurity risk assessment (V.A.2) | Residual risk after controls | Which controls exist in code and how they fail |
| SBOM and support information (V.A.4) | Every component, its support status and end-of-support date | A build that generates the component list, and the vendor facts behind it |
| Vulnerability assessment (V.A.4) | Known vulnerabilities, including CISA’s Known Exploited Vulnerabilities catalog, with risk and controls | Matching of components and versions against vulnerability sources |
| Unresolved anomalies (V.A.5) | Open bugs assessed for security impact | The bug list with a security review of each |
| Security controls and requirements (V.B.1, Appendix 1) | Requirements and acceptance criteria per control category | Security requirements in the requirements set, each with a test |
| Architecture views (V.B.2, Appendix 2) | Global system, multi-patient harm, updateability and patchability, and security use case views | Diagrams and protocol details of every path that carries code, data or commands |
| Testing (V.C) | Evidence that controls work, plus vulnerability and penetration testing | Test reports, fuzzing results, static analysis output |
| Labeling (VI.A) and management plan (VI.B) | What users are told; how vulnerabilities are handled after launch | Update procedure, logging behavior, end-of-support facts |
The threat model has a useful stated assumption. FDA recommends assuming that “hospital networks are inherently hostile” and that “an adversary controls the network with the ability to alter, drop, and replay packets.” For a Bluetooth LE patch, read that as: the phone and its radio link are hostile, and the firmware must protect itself regardless. The threat model should also cover risks from “supply chain, manufacturing, deployment, interoperation with other devices, maintenance/update activities, and decommission activities.”
Two of the four architecture views are easy to underestimate. The multi-patient harm view asks how the system defends against attacks “with the potential to harm multiple patients”, and FDA’s example is a ward of bedside monitors all restarting at once. In firmware, the usual culprits are a key shared by every device, a fleet-wide command without per-device authorization, and an update that reaches every unit at the same moment. The security use case views cover operational states such as “power on, standby, transition states” and clinical states such as programming, alarming and delivering therapy, so a boot-time or low-battery path needs the same scrutiny as normal operation.
Open bugs count as well. The guidance warns that an anomaly judged acceptable because it appears only sporadically may look different once an attacker learns to trigger it on purpose, “to manifest the anomaly’s impact continuously.” A firmware bug list reviewed only for clinical impact will miss that.
FDA also recommends discussing cybersecurity design early through its Q-Submission process. For a startup, an early pre-submission is cheaper than redesigning a boot chain after FDA asks questions about it.
If your first filing is a clinical study
Some startups meet FDA first through an Investigational Device Exemption (IDE) for a clinical study, not a marketing submission. The guidance covers IDEs too, with a lighter set. Its summary table marks as recommended for an IDE the SBOM; the global system, multi-patient harm and updateability and patchability views; security use case views for functions with safety risks; the security requirements; and labeling, including cybersecurity risks in the informed consent form. Threat models, risk assessments and test reports “could be helpful to submit, but not specifically recommended.”
So even a study unit needs a working, documented update path and an SBOM. Building them for the study is cheaper than retrofitting them before the marketing submission.
Security controls in firmware: the eight categories
FDA lists eight categories of security controls and recommends documenting requirements and acceptance criteria for each, then showing that each was implemented and tested. The categories are authentication; authorization; cryptography; code, data and execution integrity; confidentiality; event detection and logging; resiliency and recovery; and updatability and patchability (titled “Firmware and Software Updates” in Appendix 1). Most of Appendix 1 reads like a firmware design review.
Authentication and authorization
Use cryptographically strong authentication where “the authentication functionality resides on the device,” and “require authentication, and authorization in certain instances, before permitting software or firmware updates.” FDA rules out passwords that are “hardcoded, default, easily guessed, or easily compromised”, including passwords “that are the same for each device”, and asks for anti-replay measures such as nonces on potentially harmful commands.
One line deserves a place on every firmware team’s wall: “Do not rely on cyclic redundancy checks (CRCs) as security controls.” A CRC catches noise, not attackers. For authorization, FDA recommends that devices “deny by default”, for example rejecting all unauthorized incoming TCP, USB, Bluetooth and serial connections.
Cryptography and integrity
FDA recommends current NIST-recommended cryptography, no deprecated algorithms, and an architecture where “the full compromise of any single device” cannot reveal keys for other devices. It names two common shortcuts to avoid: master keys stored on the device, and keys derived from serial numbers, which “may be disclosed by patients seeking additional information on their device.”
For code integrity the guidance is specific. Verify signatures or MACs on firmware, “version numbers, and other metadata”; allow installation only of cryptographically authenticated updates; validate software “prior to execution”; and “disable or otherwise restrict unauthorized access to all test and debug ports (e.g., JTAG, UART) prior to delivering products.” On downgrades: “Do not allow downgrades, or version rollbacks, unless absolutely necessary for safety reasons, and log and document the event.”
For data integrity, validate that external data is well formed and within safe ranges, protect safety-critical configuration, and review parsing code with static analysis, dynamic analysis and manual review.
Bluetooth LE is not your security boundary
The guidance addresses common wireless protocols directly. When devices use technology such as Bluetooth LE for interoperability, FDA asks manufacturers to assess whether added security controls are needed on top of the protocol, “to protect against risks if vulnerabilities in the Bluetooth Low Energy protocol or supporting technology are discovered.” Its background section names URGENT/11 and SweynTooth as vulnerabilities in commonly used third-party components that raised potential safety concerns across many kinds of devices.
For a patch or wearable, that argues for authenticating sensitive commands and update requests at the application layer, with per-device keys, so that a flaw in the radio stack or a weak pairing mode does not hand an attacker control of the device.
Logging, resilience and updates
For embedded operating systems, the guidance says antivirus “is generally not needed unless a particular risk or threat is identified that would not be addressed by other expected security controls.” It does expect devices to detect, log and timestamp security events, to support forensic capture, and to notify users of anomalous behavior. It also suggests designing devices that can produce an SBOM in machine-readable form.
Resilience means protecting critical functions even when the device is partly compromised, recovering a trusted default configuration through an authenticated user, and stating how the device behaves when communication is disrupted, including “disruption of significant duration.”
The updates category reads like a hardware requirement. FDA expects the need for patches to “likely necessitate the need for additional storage space and processing resources”, asks manufacturers to consider “which phase of the update process the interruption or failure occurs”, and recommends security patches “independent of regular feature update cycles.” It also asks you to preserve “full build environments”, test suites, development kits and debuggers, and to keep third-party licenses for the supported life of the device.
| Control category | Firmware implementation | Evidence that it works |
|---|---|---|
| Authentication | Signed images; authenticated update and therapy commands; per-device credentials | Tests with unsigned, wrongly signed and replayed commands |
| Authorization | Deny-by-default command handlers; roles for patient, clinician and service | Tests of each command from each role |
| Cryptography | Standard algorithms; no shared master key; keys not derived from serials | Key generation and storage description; review record |
| Integrity | Secure boot, signed version metadata, anti-rollback, debug ports locked | Downgrade attempt refused; debug access attempt refused |
| Event detection and logging | Security event log with timestamps; reporting to the cloud | Log samples from triggered events |
| Resiliency and recovery | Spare image slot, revert on failed self-test, safe behavior when the link drops | Power-loss and link-loss tests during update |
| Updates | Flash and RAM budget for future patches; security-only release branch; archived toolchain | Flash budget record; reproduced build from the archive |
The secure boot checklist turns the integrity rows into pass or fail tests, and firmware security describes how we build them.
The update path FDA wants to see
Section 524B(b)(2) requires you to make updates and patches available on two clocks: “a reasonably justified regular cycle” for known unacceptable vulnerabilities, and “as soon as possible out of cycle” for critical vulnerabilities that could cause uncontrolled risks. The firmware therefore needs an update path that is secure, recoverable and fast enough to meet the second clock.
The guidance explains both clocks. The justification for the regular cycle belongs in your cybersecurity management plan, and its length depends mainly on risk. FDA’s own example contrasts “an interconnected thermometer” with “an interconnected surgery robot”, while warning that a low-risk device can still be a stepping stone to other devices in the same environment. For 524B purposes, FDA treats an update or patch as “an action that modifies device code to address a cyber risk.”
Fielded versions matter too. The guidance notes that “if an update is not applied automatically for all fielded devices, then there will likely be different risk profiles for the differing software configurations”, and vulnerabilities should be assessed for every version still in use. Firmware that reports its exact version and image hash makes that assessment possible.
FDA recommends an updateability and patchability view in every submission. It should describe the whole process by which updates and patches are deployed to the device, from your servers to the device itself, including technology “that the device manufacturer does not control”, such as a patient’s phone or a hospital network. Figure 2 sketches what that view could look like for the fictional JV-2 patch.
Appendix 2 says what each path in such a view should document. That includes protocol names and versions, whether the path carries data, code or commands, authentication mechanisms with algorithm and key length, “details of cryptographic protection for firmware and software updates”, and behavior in unexpected circumstances, such as “termination of a connection in the middle of a data transfer.”
In firmware terms, that is a signed image whose version sits inside the signed metadata, a bootloader that refuses older images, a spare slot so a failed install can revert, and a self-test before the new image is confirmed. For a medical device, add one rule FDA does not spell out but your risk analysis will likely demand: the device installs and resets only in a safe state, never during therapy or an active measurement. What it takes to ship firmware you can safely update walks the whole chain, and rollback versus anti-rollback covers the version policy that reconciles recovery with FDA’s downgrade rule.
The management plan also asks for “patching capability (i.e., rate at which update can be delivered to devices).” A staged rollout gives you that number and limits how many patients a bad release can reach; see staged firmware rollouts.
The SBOM FDA asks for, and what firmware makes hard
Section 524B(b)(3) requires an SBOM “including commercial, open-source, and off-the-shelf software components.” The guidance asks for a machine-readable SBOM meeting the NTIA baseline attributes and, for each component, two extra facts: its level of support (actively maintained, no longer maintained or abandoned) and its end-of-support date.
The guidance cites the NTIA’s October 2021 baseline attributes. On July 29, 2026, CISA, NSA, FBI and international partners published the 2026 Minimum Elements for an SBOM, which “updates and replaces” the 2021 NTIA minimum elements. As of November 2026 the FDA guidance text had not been revised to cite it, so generate SBOMs your tools can map to both and ask your regulatory lead which to reference.
Firmware makes three parts of this harder than for web software:
- Vendor SDKs and binary blobs. A Bluetooth LE controller often ships as a precompiled library. It still goes in the SBOM, with whatever version and supplier data you can get.
- Copied-in code. Drivers and libraries pasted into the tree years ago have no package manager entry. Someone has to find and record them.
- Support dates. Many embedded vendors publish no end-of-support date. The guidance allows a justification when information cannot be provided, so record “unknown” with the reason instead of inventing a date.
For each known vulnerability in those components, including those in CISA’s Known Exploited Vulnerabilities catalog, FDA wants to know how you found it, a safety and security risk assessment, and the controls that address it. Its FAQ is direct about one shortcut: asked whether VEX (a vulnerability exploitability statement) satisfies the vulnerability analysis, FDA answers “No.” VEX can carry the result of the assessment, but it does not replace it.
Example (fictional). The JV-2 SBOM lists the founders’ application code, Zephyr 3.7 LTS, the MCUboot bootloader, the vendor’s hardware abstraction layer, the Bluetooth LE controller library supplied as a binary, and a cryptography library. For Zephyr 3.7, the support entry uses the end-of-life date the Zephyr Project publishes for that long-term support release, July 27, 2029 (Zephyr releases). For the controller binary, the vendor gives no end-of-support date, so the entry reads “unknown”, with a note that the team asked in writing and will update the SBOM when the vendor answers.
Responsibility for third-party code stays with you. Asked whether suppliers of off-the-shelf software are responsible for its secure use in devices, FDA’s FAQ answers that the device manufacturer “chooses to use OTS software, thus bearing responsibility for the security as well as the safe and effective performance of the medical device,” and points to cybersecurity agreements with suppliers as part of purchasing controls. The guidance adds that purchasing controls may include acquiring the source code if a supplier’s software reaches end of support before the device does.
The guidance also asks manufacturers to keep “custodial control of device source code” over the device lifecycle, for example through escrow or backups, and to explain how third-party components “could be updated or replaced if support ends.” For a startup using a contractor, that means the repository, build instructions and signing setup sit under the startup’s control from the first commit. The mechanics of generating and maintaining a firmware SBOM are in SBOMs for embedded firmware.
Testing evidence the firmware team has to produce
FDA does not test your device. Its FAQ says cybersecurity testing, “including security requirements testing, vulnerability assessment, and penetration testing, is the responsibility of the medical device manufacturer,” and FDA reviews the results. The guidance lists what it recommends you submit.
| Test type in the guidance | What it looks like for firmware |
|---|---|
| Security requirements testing, with boundary analysis | One test per security requirement, including the limits of each input |
| Threat mitigation testing | Tests that each control in the architecture views stops the threat it was designed for |
| Abuse or misuse cases, malformed and unexpected inputs, fuzz testing | Fuzzing Bluetooth LE characteristic handlers, command parsers and the update agent’s image header parsing |
| Attack surface analysis and vulnerability chaining | A list of every interface, including unused ones, and how two weaknesses could combine |
| Known vulnerability scanning and software composition analysis of binaries | Scanning the built image, not only the source tree |
| Static and dynamic analysis, including hardcoded credentials | Static analysis in CI and a search of the image for keys and passwords |
| Penetration testing | A report stating the testers’ independence, scope, duration, methods and findings |
For any tool, the guidance suggests recording its name, version and settings. For any finding you defer, it asks for “rationales for not implementing or deferring” and plans for the future release, including whether devices released in the meantime will receive the fix and “how long it will take the update to reach the devices.” After release, it recommends testing “at regular intervals commensurate with the risk (e.g., annually).”
On independence, FDA asks who performed each test and how independent they were from the developers, and notes that third parties may sometimes be needed. A small team should plan for an outside penetration test before the submission, not after a reviewer’s question.
Most of the functional tests above can run automatically on every build. Firmware testing with CI and hardware-in-the-loop rigs explains how to set that up, and the firmware release checklist holds the update tests every release should pass.
Labeling and the postmarket plan
Two parts of the submission face outward: labeling, which tells users how to operate the device securely, and the cybersecurity management plan, which tells FDA how you will handle vulnerabilities after launch. Both depend on firmware features that are easy to leave out.
From the guidance’s labeling list, the items that need firmware support are:
- “A description of systematic procedures for users to download version-identifiable manufacturer-authorized software and firmware, including a description of how users will know when software is available.”
- How the device responds to security events, “including notification to the user and logging of relevant information.”
- Backup and restore of authenticated configurations, and the secure configuration of shipped devices.
- End-of-support information: if a device stays in service after support ends, the manufacturer should have “a pre-established and pre-communicated process for transferring the risks.”
- How to decommission the device by “sanitizing the product of sensitive, confidential, and proprietary data and software.”
- An SBOM made available to users “on a continuous basis”, in machine-readable form.
The cybersecurity management plan, which also serves as the 524B(b)(1) plan for cyber devices, should name the people responsible, the sources and frequency of vulnerability monitoring, how you will handle items in CISA’s Known Exploited Vulnerabilities catalog, periodic security testing, the timeline to develop and release patches, update processes, patching capability, your coordinated vulnerability disclosure process, and how you will tell customers about fixes.
The firmware underneath that plan needs to report its version, record security events and resets with enough context to investigate, and support a security-only release from a maintained branch. Field crash diagnostics covers what devices should capture, and the update policy worksheet records who may ship, to which devices, and how to halt.
Which firmware changes bring section 524B back
Section 524B applies again whenever a change to a cyber device needs a new premarket submission. FDA’s FAQ says the law applies to the new submission even if the device was first authorized before March 29, 2023. The guidance then sorts changes by their effect on cybersecurity.
Changes that “may impact cybersecurity and may require premarket submission” include “changes to authentication or encryption algorithms, new connectivity features, or changing software update process/mechanisms.” For those, the full Section VII documentation applies. Changes unlikely to affect cybersecurity, such as materials, sterilization or an algorithm change without architecture changes, carry a lighter set: a reference to the existing 524B(b)(1) plan, a statement on critical vulnerabilities with uncontrolled risks, and an SBOM.
On patches, FDA’s FAQ says that changes “made solely to strengthen cybersecurity and do not have any other impact on the device likely will not require a premarket submission,” based on the manufacturer’s risk-based assessment. It also says the manufacturer is responsible for validating all software changes, including security fixes.
The firmware lesson is to get three things right before the first submission, because changing them later is the kind of change FDA lists:
- The update mechanism. Choose the bootloader, slot layout, signing scheme and transport you can live with for the device’s life.
- The cryptography. FDA recommends protocols “that permit negotiated parameters/versions”. An update format that names its signature algorithm lets you move to a stronger one without redesigning the path.
- The connectivity. Adding a radio or a cloud connection later is listed as a change that may need a submission.
Leave room for growth: spare flash for larger images, space for more than one verification key, and a version field that can carry a security counter. Key custody and rotation are covered in firmware signing key management.
A firmware readiness checklist for a first submission
Before your regulatory lead starts writing, the firmware side should be able to hand over the following. Each item is a firmware deliverable with a record behind it, not a promise about review outcomes.
Design
- A cyber device determination with the reasoning: software present, every connection path listed, including service and factory connectors.
- A threat model of the whole system as built, with the hostile-network assumption stated.
- Security requirements in all eight control categories, each with acceptance criteria and a linked test.
Update path
- Signed images with version and security counter inside the signed metadata, verified by the bootloader before every run.
- Downgrades refused, with any safety exception logged and documented.
- Recovery from interrupted and failed updates, tested at each phase of the update.
- An updateability and patchability view of the full path, including the parts you do not control.
Components and testing
- An SBOM generated by the build, with support status and end-of-support dates or a written justification for unknowns.
- A vulnerability assessment of known issues, not a VEX alone.
- Fuzzing, static analysis, composition analysis of the built image, and a penetration test report with tester independence stated.
After launch
- Version, security event and reset reporting from the field.
- A security-only release branch and an archived, reproducible build environment.
- Debug ports locked on production units, and a decommissioning procedure that erases data and keys.
Medical devices describes the firmware and records we build for health and medical device startups, and our security page explains how we handle client code and keys. Work starts with a firmware review: we read the schematics, the current code, the flash layout and the update plan, and deliver a written plan with the risks and the order of work. Your regulatory lead decides what goes into the submission. If you are planning a submission for a connected device, request a firmware review.
